Metallic phase transition metal dichalcogenide quantum dots showing different optical charge excitation and decay pathways

The charge excitation and decay pathways of two-dimensional heteroatomic quantum dots (QDs) are affected by the quantum confinement effect, bandgap structure and strong exciton binding energy. Recently, semiconducting transition metal dichalcogenides (TMDs) have been intensively studied; however, the charge dynamics of metallic phase QDs (mQDs) of TMDs remain relatively unknown. Herein, we investigate the photophysical properties of TMD-mQDs of two sizes, where the TMD-mQDs show different charge excitation and decay pathways that are mainly ascribed to the defect states and valence band splitting, resulting in a large Stokes shift and two excitation bands for maximum photoluminescence (PL). Interestingly, the dominant excitation band redshifts as the size increases, and the time-resolved PL peak redshifts at an excitation wavelength of 266 nm in the smaller QDs. Additionally, the lifetime is shortened in the larger QDs. From the structural and theoretical analysis, we discuss that the charge decay pathway in the smaller QDs is predominantly affected by edge oxidation, whereas the vacancies play an important role in the larger QDs. Metallic phase transition metal dichalcogenides quantum dots show different pathways of optical charge excitation and decay according to the size and sort of defects, resulting into the large Stoke shift, two bands for charge excitation, and TRPL peak shift. This result is mainly ascribed to the valance band splitting and the emerging defect states originated from atomic vacancy of basal plane and edge oxidation.

The spectroscopic results for metallic TMDs show additional Raman modes below 200 cm −1 , a monotonic change in UV-visible absorption, and a shift in photoelectron peaks 6,20,21 . TMD-mQDs that are electrochemically, hydrothermally, or mechanically fabricated in situ with a structural phase transition have distinctive advantages, such as high charge transfer and efficient catalytic properties 4,22 . However, although the photophysical properties of metallic TMD-QDs are expected to be advantageous, most studies have focused on semiconducting TMD-QDs. This is mainly due to the various synthetic routes and particular phase engineering related to the metastable metallic phase of TMDs 6,7,21,23,24 . Phase engineering for metallic phase TMDs has been developed via methods such as intercalation of alkali metals, electron-beam irradiation, and plasmonic hot electron injection, among others, including direct synthesis 6,25,26 . Exfoliation by Li or K intercalation is the most preferred method for the synthesis of metallic phase TMDs 5,7,20,21 , where high quality and phase stabilization are achieved simultaneously. Moreover, the phase stability is enhanced by lithiation, doping, vacancy defects, and lattice strain 7,21,27,28 .
In this study, we report an in-depth investigation of the photophysical properties of metallic phase (1T/1T′) TMD (MX 2 , M = Mo, W; X = S, Se)-QDs synthesized using the K and Na intercalation method, in which the TMD-mQDs were separated into two groups of different sizes (mQDs-I:~3 nm, mQDs-II:~12 nm). The metallic phase and structural properties were confirmed by high-resolution transmission electron microscopy (HR-TEM) showing an octahedral crystalline structure, novel Raman vibrational modes below 200 cm −1 , and redshifted X-ray photoelectron spectroscopy (XPS) spectra. For the photophysical properties, various spectroscopic methods were employed, including excitation-dependent PL (PLE) with two resonance excitation bands and peak shifted timeresolved PL (TRPL) spectroscopy. Density functional theory (DFT) calculations revealed that the oxidation defects at the edge structure and vacancies on the basal plane play a role in controlling the electronic structure and exciton dynamics. Based on these analyses, a photoelectronic model of TMD-mQDs is suggested.

Synthesis and characterization
For years, we have developed the process of cointercalation and exfoliation of TMDs using alkali metal-organic compounds, such as potassium sodium tartrate tetrahydrate salt, as shown in Fig. 1 (see details in the Experimental section and Supplementary Information) 5 . In this study, we adopt this method to synthesize metallic phase TMD (MoS 2 , MoSe 2 , WS 2 , WSe 2 )-QDs. The as-prepared TMD-salt compounds were instantly dispersed in water and sonicated for a few minutes, followed by filtration and dialysis. During the sonication process, TMD-QDs were explosively synthesized. Atomic force microscopy (AFM) and high-resolution transmission electron microscopy (HRTEM) revealed that the TMD-QDs separated into two groups have sizes of~3 nm (QDs-I) and~12 nm (QDs-II; Fig. 2a). The AFM measurement yielded a QD thickness of 1-2 nm, regardless of the lateral size, implying that most QDs are mono-or bilayers ( Supplementary Fig. S1). Figure 2b shows digital photographs of QDs-I and QDs-II dispersed in water at a concentration of 0.1 mg/ml, where QDs-I are significantly brighter and closer to the blue color compared to QDs-II under UV illumination. The dispersion stability is longer than 1 month, which is attributed to the partially negatively charged QDs, as examined by the zeta potential ( Supplementary Fig. S2). Furthermore, HRTEM images directly demonstrate the lateral size and circular structure of the QDs ( Fig. 2c and Supplementary Fig. S3). The analysis of magnified HRTEM images (inset of Fig. 2c) shows that the lattice constants of the (100) plane are 0.25, 0.23, 0.2, and 0.24 nm for MoS 2 , MoSe 2 , WS 2 , and WSe 2 , respectively, which are slightly smaller than those reported for 2H semiconducting TMD layers 11,18,24,29 . The HRTEM image of QD-II shows point defects on the basal plane that are most likely related to chalcogen or metal atom vacancies (Fig. 2d, top) 1,30 . The corresponding fast Fourier transform pattern demonstrates the octahedral crystalline structure of QDs typically observed in metallic phase TMDs (Fig. 2d, bottom). The metallic Novel vibrational modes assigned to J 1 (146.5 cm −1 ), J 2 (235 cm −1 ), and J 3 (335.5 cm −1 ) bands appear in both Raman spectra of MoS 2 QDs-I and QDs-II, which are indicative of the distinct Raman signature of metallic phase MoS 2 ( Fig. 2e) 7,31 . Additionally, WSe 2 QDs show distinct vibrational modes below 225 cm −1 related to the J 1 , J 2 , and J 3 transitions, while the absence of a B 1 2g peak at 304 cm −1 indicates a single layer similar to WSe 2 nanosheets (Fig. 2f) 24 . With the similar tendencies in all WS 2 and WSe 2 -QDs 32,33 , novel vibrational modes below 200 cm −1 are noted, suggesting that the as-prepared TMD-QDs are in the metallic phase ( Fig. S4a, b). In the XPS spectra of the four different TMDs-QDs (Fig. 2g, h and Supplementary Fig. S4c, d), the transition metal and chalcogen peaks are shifted to lower binding energies and broadened with additional peaks compared to those of semiconducting 2H phase TMDs 18,34 . In the XPS spectra of Mo ( Fig. 2g and Fig. S4c), the Mo3d 5/2 peak is shifted 1 eV lower compared to that measured from 2H semiconducting MoX 2 . In addition, separated W4f peaks clearly appear in the W spectra (Fig. 2h), and the assigned 1T W4f 5/2 peak is~1 eV lower than the 2H W4f 5/2 peak (Fig. S4d). However, the characteristic in-plane E 1 2g and/or out-of-plane (A 1g ) vibrational modes in the Raman spectra and the peak shift of the metal atom to a lower binding energy in the XPS spectra indicate that the asprepared TMD-QDs could contain some portion of the polymorph structure, although the metallic phases (1T/1T′) are still dominant.

Electronic and optical properties
The as-prepared TMD-mQDs-I and -mQDs-II were investigated by UV-visible (UV-vis) absorbance and PLE (i.e., excitation wavelength (λ Ex )-dependent emission wavelength (λ Em )) measurements to determine their   electronic and optical properties (Fig. 3). Figure 3a clearly shows monotonically decreasing UV-vis absorbance spectra in both mQDs-I (left) and mQDs-II (right), indicating the metallic properties of the QDs. The Tauc plot applied for the direct transition bandgap (E og ; inset of Fig.  3a) provides estimated values of E og = 4.14-4.20 eV, except for MoSe 2 , with values of~3.85 eV in mQDs-I and 3.88-3.93 eV in mQDs-II, although there is a deviation depending on the stoichiometry of the TMDs. The E og values larger than those of bulk or single-layer TMDs are assigned to an excitonic feature of the mQDs reflecting a strong quantum confinement effect 2,11,13,16 , whereas the difference between mQDs-I and mQDs-II originates from the size difference. Figure 3b shows the PL spectra of mQDs-I excited at λ Ex~2 60 nm, which is >E og , and those of mQDs-II excited at 360 nm, which is slightly <E og . The spectral shapes are similar in each group but have different emission peaks (λ peak ) and linewidths (FWHM) (mQDs-I: λ peak = 427 ± 3 nm (2.90 eV), FWHM = 143 nm; mQDs-II: 464 ± 7 nm (2.67 eV), FWHM = 184 nm). The large Stokes shifts (~1.38 eV and~0.78 eV) of the mQDs are noteworthy and ascribed to the large valance band splitting due to the strong spin-valley coupling, midgap state caused by defects, and enormously large exciton binding energy (>500 meV) in the mQDs in addition to vibrational relaxation and trion formation 18,35,36 . Figure 2c-f shows the PLE intensity of mQDs-I and -II of MoS 2 and WSe 2 , where the measured λ Em varies from 380 to 640 nm. Interestingly, in mQDs-I, two persistent intensity maximum peaks clearly appear at every λ Em (Fig. 3c, d). The former is located at λ Ex = 240-260 nm (A-band), and the latter is located at λ Ex = 290-375 nm (B-band). These two excitation bands are consistently observed in another mQDs-I (MoSe 2 and WS 2 , Supplementary Fig. S5a, b). The λ Ex longer than 375 nm is related to the midgap states originating from defects 18 . The luminescence intensity peak in the A-band is independent of the stoichiometry of the TMDs for all measured λ Em , whereas the B-band peak is redshifted and broadened as λ Em increases. However, it is also weakly correlated with the stoichiometry The mQDs-I were excited at λ Ex~2 60 nm, which was greater than E og , and the mQDs-II were excited at 360 nm, which was slightly less than E og. The insets in a are Tauc plots ((αhν) 2 vs. hv) for the optical bandgap. c-f PLE intensity of mQDs-I and -II of MoS 2 (c, d) and WSe 2 (e, f), where the measured λ Em varies from 380 to 640 nm. The emission wavelength (λ Em ) was measured from 380 nm to 560 nm. g-j. 2D contour plots of λ Ex vs. λ Em for Fig. 3c- Fig. S5c, d). Interestingly, in mQDs-I, the PL intensity peak in the A-band is higher than that in the Bband, whereas it is similar to or smaller than that in the Bband in mQDs-II. For example, in WSe 2 mQDs-II, the PL intensity in the A-band becomes like or smaller than that in the B-band (Fig. 3f). The two bands are certainly not related to the neutral free excitons or bound excitons of TMDs because of the monotonically decreasing UV absorption and single peak of the PL spectrum. This is probably due to the band splitting by strong spin-valley coupling and the quantum size effect because the ΔE between the two PLE peaks is~0.7-0.9 eV, which is like the valence band splitting in semiconducting TMD-QDs 18 .
To date, this feature has not been reported for semiconducting TMD-QDs, which will be discussed more later. Two-dimensional contour plots for the correlation between the excitation and emission wavelengths are shown in Fig. 3g-j and Supplementary Fig. S6e-h. Two regions of luminescence maximum intensity corresponding to the PLE spectra are evidently shown. The point of maximum intensity in mQDs-I oscillates slightly as λ Ex increases from 250 to 340 nm, although it is gradually redshifted as λ Ex increases beyond 300 nm, which is usually observed in QDs. In contrast, the oscillation propensity of mQDs-II is weakened, while the highest intensity area moves into the B-band region.

Exciton dynamics
To further investigate the exciton dynamics of the TMD-mQDs, TRPL measurements were performed at two λ Ex . As shown in Fig. 4a, b, the TRPL spectra of WSe 2 mQDs-I at λ Ex = 266 nm show that the PL peak gradually redshifts from~410 to~470 nm as the delay time increases, whereas in WSe 2 mQDs-II, the redshift of the PL peak is trivial even after a delay time of 30 ns. Quantitative analysis of the peak shift at λ Ex = 266 nm reveals that while the peak of mQDs-I redshifts to above 200 meV, it is <100 meV in mQDs-II (Fig. 4e). In comparison, the peak shift of the TRPL spectra at λ Ex = 360 nm in both mQDs-I and -II is much less than that in WSe 2 mQDs-I excited at λ Ex = 266 nm (Fig. 4c, d). The TRPL peak shift at λ Ex = 360 nm is <100 meV in both mQDs-I and -II (Fig. 4f). This feature is analogously observed in all other TMD-mQDs ( Supplementary Fig.  S7). This suggests that some excitons formed at λ Ex = 266 nm on mQDs-I steadily transit to the lower energy states together with an emissive decay, whereas excitons formed at λ Ex = 360 nm on mQDs-II are likely to predominantly decay at the occurring sites. This speculation is indirectly supported by the PL decay dynamics.
Although all exciton decay profiles from mQDs-I and -II are best fitted by a multiexponential function, the profiles from mQDs-II are closer to the monoexponential decay line (Fig. 4g). The average PL lifetime of mQDs-I is 3.94 ± 0.12 ns at λ Em = 420 nm, which is longer than the value of 2.98 ± 0.11 ns for mQDs-II at λ Em = 460 nm ( Fig. 4h and Supplementary Fig. S8). This indicates that the exciton decay dynamics are different between mQDs-I and -II, which can be ascribed to the different emission sites originating from different defects on the mQDs.

Theoretical calculation
To prove our speculation that the electronic structure is affected by defects on TMD-mQDs, first-principles calculations within the density functional theory (DFT) framework were performed on the representative structural model of MoS 2 . Figure 5a shows the variation in the bandgap of MoS 2 mQDs without defects with the QD size, where the bandgap increases as the size decreases due to the quantum confinement effect. However, during the synthetic process, the formation of vacancy defects on the basal plane and oxidation of edge-exposed atoms are inevitable. Figure 5b schematically illustrates the chalcogen vacancy (top) and oxygen defect (bottom) of the MoS 2 structure. Structural defects are assumed based on XPS analysis ( Supplementary Fig. S5). When considering such defects, the electronic bandgap of MoS 2 is narrower than that of the pristine structure (Fig. 5c). Unexpectedly, with the band structure calculations ( Supplementary Fig. S9), the vacancy defect on the basal plane conspicuously leads to a bandgap narrowing up to 0.2-0.4 eV compared to that of pristine MoS 2 , which is a significantly larger variation than that induced by oxygen defects on the edge structure (< 0.1 eV) 37 . Furthermore, the energy difference (0.2-0.3 eV) between them corresponds well to the difference (~0.3 eV) between the PL peaks of mQDs-I and mQDs-II (Fig. 3b).

Discussion
Herein, the photophysical properties of metallic phase TMD-QDs are reported by comparing mQDs-I (~3 nm) and mQDs-II (~12 nm) of two sizes. Based on the above results, a model for possible optical excitation and relaxation pathways in mQDs-I and mQDs-II is schematically illustrated in Fig. 6a, b. Because the ΔE (~0.8 eV) between two PLE excitation bands (Fig. 3c-f) corresponds well to that of valence band splitting in TMD-QDs 18 , the system is assumed to be resonantly excited from the valence band splitting states to the conduction band (Sl and higher energy states) and lower defect states. Due to the metallic phase of QDs, strongly coupled excitonplasmon polaritons might occur 38 , where the resonance frequency varies with the size of mQDs. This, along with the defect-induced anisotropic confined excitons 12 , may cause oscillation of the PLE intensity peak (Fig. 3g-j) and variation of the relative quantum efficiency in mQDs-I and mQDs-II after considering UV absorption (Supplementary Fig. S10). Additionally, as the edge oxidation of mQDs-I with a size comparable to the exciton Bohr radius has a dominant effect on the electronic structure (Fig. 6a), which is in contrast to mQDs-II mainly affected by vacancy defects (Fig. 6b), the radiative decay site of excitons is different from that in mQDs-II, which results in the difference in the energy (~0.3 eV) and lifetime (~1 ns) of PL. This is supported by structural analysis (Fig. 2 and Supplementary Figs. S3-5) and theoretical DFT calculations (Fig. 5). In addition, the vacancy and edge oxidation defects inducing substantial electron density localization cause a large Stokes shift (Fig. 3b) 35 .
In summary, the electronic and optical properties of metallic phase TMD-QDs (mQDs) revealed that the exciton quantum dynamics are affected by the quantum confinement effect, band splitting due to strong spinvalley coupling, and defects. Two resonance excitation bands and oscillation of the PLE intensity peak of TMD-mQDs are first reported, which have not yet been reported in semiconducting TMD-QDs. When the size of mQDs is close to the exciton Bohr radius, edge oxidation has a dominant effect on the electronic structure; however, if the size is sufficiently larger than the exciton Bohr radius, then the vacancy on the basal plane acts as the main site determining the electronic structure and exciton dynamics. Our findings expand the general understanding of the nonlinear photophysics of two-dimensional heteroatomic structured QDs, which is essential for the potential applications of excitons-plasmons induced by light-matter interactions in optoelectronics and biooptoelectronics.

Metallic phase TMD-QDs
Potassium sodium tartrate was used to intercalate Li and K into the TMDs at low temperature. The metalintercalated compounds of TMDs were reacted in the autoclave vessel at 250°C for 12 h and then instantly exfoliated in water with sonication. The synthesized mQDs were separated into two sizes through filtration and dialysis5.

Density functional calculations
DFT calculations were performed using the generalized gradient approximation (GGA) for the exchange-correlation potential and projector augmented wave potentials, as implemented in the VASP code. The wave functions were expanded in plane waves up to an energy cutoff of 400 eV. We employed a supercell geometry with a vacuum region of more than 8 Å to prohibit interactions between adjacent supercells (more details are given in the Supplementary  Information).